Can a machine learn to write for New Yorker Magazine?

The safety of any new technology often hinges on how it’s regulated. If machines can learn to think for themselves, that might be a concern. But if we really want to replicate human intelligence—as most of us want to—there are several directions that researchers might explore.

The italicized paragraph above was written by a machine. Here’s the story by John Seabrook in The New Yorker:

Machine translation, an enduring dream of A.I. researchers, was, until three years ago, too error-prone to do much more than approximate the meaning of words in another language. Since switching to neural machine translation, in 2016, Google Translate has begun to replace human translators in certain domains, like medicine. A recent study published in Annals of Internal Medicine found Google Translate accurate enough to rely on in translating non-English medical studies into English for the systematic reviews that health-care decisions are based on.

Ilya Sutskever, OpenAI’s chief scientist, is, at thirty-three, one of the most highly regarded of the younger researchers in A.I. When we met, he was wearing a T-shirt that said “The Future Will Not Be Supervised.” Supervised learning, which used to be the way neural nets were trained, involved labelling the training data—a labor-intensive process. In unsupervised learning, no labelling is required, which makes the method scalable. Instead of learning to identify cats from pictures labelled “cat,” for example, the machine learns to recognize feline pixel patterns, through trial and error.

More here.

Like what you're reading? Don't keep it to yourself!
Share on Facebook
Facebook
Tweet about this on Twitter
Twitter
Share on Reddit
Reddit
Share on LinkedIn
Linkedin
Email this to someone
email